As someone who builds models with uncertainty, you know that decision tree analysis and Monte Carlo
simulation are complementary, with each having advantages for particular problems. With DPL
you have a choice, you can decide on a per-model basis which form of analysis to use.
New Continuous Chance Nodes
To use Monte Carlo simulation in DPL, you need to create some continuous chance nodes in your model. Continuous
chance nodes are like discrete chance nodes without states. They are always initialized by
a named probability distribution. DPL draws continuous chance nodes in dark green, to distinguish them
from their discrete cousins. In the tree, DPL uses shading to suggest a continuum of branches.
Continuous and discrete chance nodes are both created in the influence diagram, and are both placed
in the decision tree to show sequence. Here are a few things to keep in mind.
How DPL is Different
- A model is continuous if it has one or more continuous chance nodes, else it is discrete.
- Run | Decision Analysis is for use with discrete models.
- Run | Monte Carlo Simulation is for use with continuous models.
- Continuous chance nodes can be freely conditioned on decisions or discrete chance nodes.
Monte Carlo simulation takes many forms. These are some of the capabilities that distinguish DPL.
A Running Example
- DPL's Monte Carlo is built on a fast tree engine, so it supports advanced outputs like the Policy Summary and Policy Tree.
- With DPL, you can convert your spreadsheet to code for a typical speed increase of 10-100 times, so you can avoid the random noise of small sample runs without paying the price in long runtimes.
- DPL supports downstream decisions, so you can perform serious real options analysis, where the exercise policy isn't an =IF statement.
- You can mix discrete and continuous elements to build asymmetric hybrid models.
To get a better idea of how DPL Monte Carlo works, take a look at the sample files delivered with
DPL -- Biz Unit Risk Analysis.da and Biz Unit DCF.xls.
The spreadsheet is set up to perform a valuation of a service business which faces significant regulatory risks.
The several models in the DPL project file take you through sensitivity analysis, probabilistic risk analysis and
the evaluation of a decision to spin off the business.